ServiceNow and IBM results trigger software stock selloff By Investing.com

This is less a single-name earnings miss than a regime shift in how software multiples should be underwritten. The market is discounting a double hit: cyclical budget caution from geopolitics now, and structural margin compression later if AI agents start unbundling seat-based pricing. That combination is toxic because it destroys the usual defense that software names get relief from recurring revenue quality during macro slowdowns.

The first-order losers are the horizontal workflow and back-office franchises most exposed to “good enough” automation. The second-order winners are not necessarily the hyperscalers themselves, but the infra layers that monetize usage, distribution, and model access rather than per-seat subscriptions; that supports relative strength in platform AI names versus application-layer SaaS. Expect enterprise buyers to stretch implementation cycles over the next 1-2 quarters, which disproportionately hits net-new ACV and elevates churn risk in smaller vendors before it shows up in headline growth.

The setup into big tech earnings is asymmetric: if the cloud/platform cohort can show AI is accretive to spend and not just cannibalistic, the entire software de-rating can pause for 2-4 weeks. If not, the selloff broadens from multiple compression to estimate cuts, which is where drawdowns get larger and faster. The key is that the market is already pricing a lot of bad news; the next leg depends on whether management teams can credibly frame AI as margin expansion rather than product substitution.

Contrarian take: the move may be too indiscriminate in the highest-quality franchises. Names with embedded mission-critical workflows, high switching costs, and strong free cash flow can absorb pricing pressure better than the market is implying, while weaker mid-cap SaaS remains vulnerable to valuation air pockets. The risk is not that software goes to zero; it is that dispersion explodes and investors stop paying for growth without proof of AI durability.